Luisa Crawford
Nov 06, 2024 16:26
NVIDIA introduces new AI and simulation tools, including the Isaac Lab framework, to enhance humanoid robot development, revealed at the Conference for Robot Learning in Munich.
NVIDIA has announced a suite of new AI and simulation tools aimed at accelerating the development of humanoid robots, according to NVIDIA. This announcement was made at the Conference for Robot Learning (CoRL) in Munich, Germany, showcasing NVIDIA’s commitment to enhancing the capabilities of robotics developers.
Revolutionizing Robot Learning with Isaac Lab
The highlight of NVIDIA’s announcement is the general availability of the NVIDIA Isaac Lab, an open-source robot learning framework. Built on the NVIDIA Omniverse platform, Isaac Lab facilitates the training of robot policies at scale, applicable to various robot embodiments such as humanoids, quadrupeds, and collaborative robots. The framework is being adopted by leading companies and research entities worldwide, including Agility Robotics and Boston Dynamics.
Project GR00T: Pushing Humanoid Development
NVIDIA also introduced Project GR00T, an initiative designed to accelerate the development of humanoid robots. This project offers six new workflows that provide comprehensive blueprints for overcoming the challenges associated with humanoid robot capabilities, such as motion generation and whole-body control. Jim Fan, NVIDIA’s senior research manager, emphasized the collaborative efforts behind Project GR00T to advance humanoid robotics globally.
Innovative Tools for World Model Development
To support the creation of AI representations of the world, NVIDIA unveiled the Cosmos tokenizer and NeMo Curator. The Cosmos tokenizer enhances video and image data processing with high-quality compression, while NeMo Curator streamlines video processing, significantly reducing data curation time. These tools are essential for developing accurate and efficient world models, crucial for robot interaction with their environments.
Commitment to the Robot Learning Community
NVIDIA’s participation at CoRL included the presentation of 23 research papers and nine workshops, addressing various aspects of robot learning. These contributions highlight breakthroughs in vision language model integration, robot navigation, and humanoid robot control. The conference served as a platform for NVIDIA to share its advancements and collaborate with the global robot learning community.
The newly released tools and frameworks are available on platforms like GitHub, with additional resources such as developer guides and tutorials accessible for those interested in exploring NVIDIA’s innovations in robot learning and simulation.
Image source: Shutterstock